Baseline Hybrid FWI Model

This project provides a baseline hybrid model for Full Waveform Inversion (FWI), combining physics-based and machine learning approaches to reconstruct subsurface images from seismic data.

Features

• Physics-based 2D wave equation forward operator
• U-Net-based neural network for inversion
• Hybrid loss: combines data and physics consistency
• Synthetic data generation for demonstration

Getting Started

1. Install dependencies:

   pip install -r requirements.txt

2. Run training:

   python train.py

3. Evaluate the model:

   python evaluate.py

File Structure

• models/unet.py: U-Net model
• models/physics.py: Physics-based forward operator
• utils.py: Data loading and synthetic data generation
• train.py: Training script
• evaluate.py: Evaluation script

Notes

• This starter uses synthetic data for demonstration. Replace with real data as needed.